Shot Segmentation and Classification in Basketball Videos

Shot is important and meaningful unit in sport videos. In this paper, we propose a shot segmentation algorithm using DBNF (the difference between neighbour frames) and DBSF (the difference between a frame and a designed start frame). Then we propose a shot classification algorithm based on dominant colours and edge image, by which the shots can be classified into court shots and non-court shots. Experiments show that proposed shot segmentation algorithm is effective to both abrupt cuts and wipes, and shot classification algorithm can extract more than 90% court shots correctly.